It is worth showing the link assembly issues with pictures rather than just words to get the message better across, so I've drawn some (not to scale, dimensions exaggerated, showing the principles).
This is the 831 assembly of the upper link to one of the rear triangle eyelets, as far as I could make it out. Grey is the aluminum link, black the carbon triangle, blue the ball bearing and pink the screw. The screw presses the inner race of the bearing against the link, so the screw and inner race are one piece with the link. There is some clearance between the screw head and the triangle so that the parts can rotate with respect to each other.
JJJ thanks for sharing. This definitely helps everyone who has not studied this linkage for hours understand the issue happening inside this pivot. I am not convinced however that it is only the linkage at fault. I believe this is part of the issue (and likely easier to correct) than the 2nd issue which I believe is the triangle being flexy on its own. If anyone has evidence to support that the triangle is appropriately stiff because they jammed enough nylon into the pivot that it stopped flexing all together (within what we know is acceptable based on other bikes) please prove me wrong. I have inserted the outer bushings by making a custom set of nylon spacers. It seems to be helping some. I also inserted some smaller “janky” inserts race but It was binding the whole pivot from moving freely which was unacceptable. Perhaps if some perfectly toleranced metal inserts where made it would work, but to JJJ points the engineering is faulty.
Why didn’t they just replicate the SC pivot? I rode with a friend yesterday that had a hightower and studied it pretty closely. The bike is very similar with the major difference being how they handled this set of bearings and pivot. Why they decided to reinvent the wheel on this is beyond me. Seems like they should have stuck with both sets of bearings in the pivot. They did so for the top bearings in the link, but decided to do their own thing for the lowers. Can only assume it was related to production efficiency or something similar.